Autonome, adaptive Bewegungskoordination von Gehmaschinen in komplexer Umgebung
Die Dissertation erscheint auch im VDI-Verlag, Düsseldorf. - 2003. Fortschrittberichte : Reihe 8 ; Nr. 995. - ISBN 3-18-399508-1
The text describes methods to enable the developed six-legged Tarry walking machines to move autonomously in unknown terrain. To facilitate this, walking patterns for undisturbed movement are adapted und recombined to deal with unstructured terrain. Walking patterns that have been computed analytically are approximated by artificial neural networks to fulfill the time constraints. The movement coordination adapts the input values of the neural networks to the recognized environment by means of artificial reflexes and so called "planning modules". Reflexes are using a simple stimulus-reaction schema, e.g. to climb over obstacles. Planning modules gather and save sensory input. This data together with the current walking parameters are used for plan ahead actions as bypassing obstacles. Experiments that have been made to check the achieved results are documented.